langchain-searchapi

An integration package connecting SearchApi.io and LangChain.

Why this package?

LangChain's original SearchApi integration lives in langchain_community (merged in 2023) but was never ported to the modern standalone package architecture. It lacks:

• Multi-engine selection at call time — the old tool only accepts a query string; agents can't switch engines dynamically
• Retriever for RAG — no way to use SearchApi results in retrieval chains
• Secret management — API key stored as plain string, visible in logs/repr
• Structured output — returns stringified results instead of typed dicts

This package (langchain-searchapi) is a modern, standalone replacement following the same architecture as langchain-tavily. It provides a Tool for agents, a Retriever for RAG, and full async support.

Installation

pip install langchain-searchapi

Setup

Set your API key as an environment variable:

export SEARCHAPI_API_KEY="your-api-key"

Get your key at searchapi.io.

Quick Start

Tool (for agents)

from langchain_searchapi import SearchApiSearch

# Default Google search
tool = SearchApiSearch()
results = tool.invoke({"query": "latest AI news"})

# Switch engines at call time
results = tool.invoke({"query": "python tutorial", "engine": "youtube"})

# Configure defaults at initialization
tool = SearchApiSearch(engine="google_news", num=5, gl="us", hl="en")
results = tool.invoke({"query": "technology"})

With a LangChain Agent

from langchain_searchapi import SearchApiSearch
from langchain_openai import ChatOpenAI
from langgraph.prebuilt import create_react_agent

llm = ChatOpenAI(model="gpt-4o")
tools = [SearchApiSearch()]

agent = create_react_agent(llm, tools)
response = agent.invoke({"messages": [("user", "What happened in tech news today?")]})

Retriever (for RAG)

from langchain_searchapi import SearchApiRetriever

retriever = SearchApiRetriever(engine="google", num=5)
docs = retriever.invoke("machine learning frameworks comparison")

for doc in docs:
    print(f"{doc.metadata['title']}: {doc.page_content}")

In a RAG Chain

from langchain_searchapi import SearchApiRetriever
from langchain_openai import ChatOpenAI
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnablePassthrough

retriever = SearchApiRetriever(num=5)
llm = ChatOpenAI(model="gpt-4o")

prompt = ChatPromptTemplate.from_template(
    "Answer the question based on the context:\n\n{context}\n\nQuestion: {question}"
)

chain = (
    {"context": retriever, "question": RunnablePassthrough()}
    | prompt
    | llm
    | StrOutputParser()
)

answer = chain.invoke("What are the best Python web frameworks in 2024?")

Supported Engines

API Reference

SearchApiSearch

Tool for use with LangChain agents. Agents can select the engine and number of results dynamically via the tool's input schema.

SearchApiRetriever

Retriever for RAG chains. Returns search results as LangChain Document objects with metadata (title, link, source, position).

Same parameters as SearchApiSearch.

SearchApiAPIWrapper

Low-level wrapper if you need direct API access. Provides raw_results(), results() (metadata-stripped), and run() (text snippets).

Migrating from langchain_community

If you're using the deprecated langchain_community.utilities.SearchApiAPIWrapper:

# Before (deprecated)
from langchain_community.utilities import SearchApiAPIWrapper
wrapper = SearchApiAPIWrapper()
result = wrapper.run("query")

# After
from langchain_searchapi import SearchApiAPIWrapper
wrapper = SearchApiAPIWrapper()
result = wrapper.run("query")

The new package is a drop-in replacement for basic usage, with additional features (multi-engine tool schema, retriever, SecretStr key management).

Related

• SearchApi.io Documentation
• LangChain Documentation
• langchain_community SearchApi (deprecated) — the old integration this package replaces

License

MIT — see LICENSE for details.